96

THEPROCEEDINGS OF

IEEE

JANUARY

for the oddlength Barkercodes sine.alternate c o l u m n s s u m to zero. This

could have beennoted in advance of computation since these codesfall in
the general class
QYz) = Q ( - 4

(4)

such that only evenpolynomials in z canbe generated by the multiplication

of (3) One way ofcompacting the multiplicationis noted in Fig. 1, where
+ and - represent + 1 and - 1. The tabular entries in
each column
are not obtained by a conventional multipIication format, but rather by
collecting multiplications from cross-term entries equally spaced on
either side of a i c o l u m n , one from H z ) and the oppositefrom Q(-z).
Since codes ofthe form(4) yield entries (f2) only, the two factor may be
entered after thecolumn summations as noted, andas a last step the entries
from like powers of (3), e.g., self-terms, are added. With short practice,
this multiplicationis at least as easy as conventional multiplication
once it
is noted that the multiplicands for a particular entry are foundon intersecting 45 diagonals. Although onehas reduced the number of required
multiplicationsin the scheme of Fig. 1 by nearly a factor of four, its real
value isthat it permits further abstraction.

LET z = ejaT
Q k )

a*(z)=otd

zbPz4

a;(
SELF
TERYS
SUM

~ 3 z - 2 ~P4

22

23 24

PP

Fig. 3. Correlation function description by checkerboard pattern change lines.

+-++-+
++--+
- -+++++
++
+

0 4.2 0 42 0 +2
12
ti -I +I -I ti

{ +I - I
I

zQz~Pz~2%*

Fig. I .

21

+++++--++-+-+
+ -+ - + + - -++ + + +

P 22

24

Fig. 2. Checkerboard code pattern.

P B PP

might appear from simple examples; and error is twice as effective as it

might bein a more
general code.
We close with the remark that, since codes of the form of (4) are
redundant past theirmiddle element, it is possible in principle to process
and generate them with delay lines which are half the length of conventional practice.
T. R. O M m
Hughes Research Labs.
Malibu, Calif.

Condation hmaionfor Barter codc of thirteen.

Thus for binary codes of the prescribed form, corresponding even

powers in [ H z ) , e<-.)] have only two possibilities: (a) [ 1, 11, or
(b) [-I, - l ] ; s ~ y , t h e o d d p o w e r s m a y o n l y b e ( a ) [+1, -11,or
(b) [ - 1, + 1 1. Thus the code dements and the multiplications may be
described in terms ofa basic pattern (we choose the(a) forms) and an inversion or change at s e l e c t e d points. Remembering that multiplicand entries are found on 45 diagonals, it follows that changes occur at every An Exact Expression for the Thermal Variation of the Emitte
entry alongthese diagonals, which
will be called change lines.
Base Voltage of Bi-Polar Transistors
For thebasic (a) patterns,Fig. 2 indicates that theunchanged code and
Over the years, a number of interesting and
useful properties concerning
its multiplicationentires (excepting the top line) form a checkerboard pat- the highly predictable nature
of the emitter-base
voltage of bi-polar transistern with desirable sidelobe levels (+ 1) everywhere. Unfortunately, the tors have been discovered. First, it was shown that relationship between
main lobe is also + 1 and the multiplication does not produce an auto- the collector current and the emitter-base voltage exactly followed the
correlation function in the previous sense. However, by introducing a diode equation over more than eight decades [l], [2]. This has been put
limited number of changesalong diagonals as noted in Fig. 3, both defects directly to use inmaking logarithmic amplifierswithextremelywide
are easily corrected. Further, with moderate care, for Barker Codes at dynamic ranges 131. Secondly, the theoretically predicted behavior of the
least, it proves easy to pair the change points, spacing them at distances emitter-basevoltage has made possible dc amplifiers withdrifts an order
equal to an odd multiple of the row spacing. Thus, equal and opposite of magnitude lower than canbe obtained with conventional techniques
141.
change pairs are denotedby the heavy vertical lines which terminate on This same theory has been used to produce an ultra-stable, temperaturethem. Since no net changes are introduced in this example, the sidelobe compensated reference element [5] as well as a low-valve current source
level is unchanged, although the changes on the center line have rendered which does not require large resistance valves, making it well suited for
all thesecode productspositive. In longer c
o
d
e
s
,of course,some unpaired integrated circuitry [6], [7]. Lastly,it has been shownthat exact temperachanges will occur and the sign of the change is noted and entered as a ture compensationof the transconductancecan be realized [8].This has a
perturbation on the checkerboardsidelobe level of + 1. Clearly, negative number of applicationsincluding a differential input stagewhere common
to positive changeshurt mostif the objective is to minhizethe maximum mode rejectionand gain bandwidthproduct must be maximized and offset
sidelobe excursion.
voltage and thermal drift must be minimized, as in a core-memory sense
Although the autocorrelations are symmetrical and one would not amplifier 191.
generally bother with the right half of the pattern, it has been included in
This letter will demonstrate another
predictable property of the emitterLine patterns are not symmetrical. Thus
Fig. 3 to illustrate that the change
base voltage in deriving an expression for the emitter-base voltage as a
in a sense one has twochances to discover an identical code.
It is possibleto manually synthesize coda of length 6fty or more by
this technique, but it is not as easy to hold down the sidelobe level as it
Manuscript received Oaobcr 13,1966.

+ +

PROCEEDINGS LETTERS

1967

97

function of temperature in terms of physical constants and the emitterbase voltage at any one temperature. Since a direct attack on this problem
quickly produces unwieldy equations, a backdoor approach, based on the
be
transconductancecompensation schemementionedpreviously,will

RL

Used.

It is shown in [9] that,assuming very high current gains and low-level

injection conditions, the voltage gain of the differential amplifier in Fig.1
will be constant with temperature for
VB =

n
-V

kT

v,,+ (n - l ) -

V,

Where
is the extrapolated energy gap for the semiconductor material at absolute zero, q is the charge of an electron, n is a constant which
depends on how the transistor is made, k is Boltmanns constant and T
is absolute temperature. The transconductance of the differential pair is

Fig. I . Temperaturecompensated differential amplifier used as basis for the derivation.

so the gain of the stage is

Hence,

In order to have temperature-stable gain, it must be true that

and

I,a T .
From Fig. 1,
Ic -

vB

vBE

- BT

This is the change in emitter base voltage caused by maintaining a constant collector current. .The complete expression forVBAT)is then

RB

where B is an arbitraryconstant. Hence,

VBE = V
- BTRB

For T = To,
Vmo = V
- BToRB,

and

The extrapolated energy gap (V,,) for silicon is 1.205V, k/q has a value of
8.66 x lo- VPC, and the constant, n, has a typical value of 1.5
for doublediffused silicontransistors.
To give some appreciation for the magnitude of the terms in (15). a
sample calculation can,be made for T0=25c, VBE0=670 mV and
T= 125C:
VBE

Substituting into (6), this becomes

v -BE

T
1 - - VB+ - v,,

( %)

To

Using (1) for VB.

v,,=

( - -3v,,+
1

-v,,,
3+ -:

With I , varying linearly with temperature,

I,

T
I,, -
TO

- 0.403 + 0.894 - 0.010- 0.006 = 0.476 V

This shows that the last two terms of (15) are relatively small, making
V,, nearly a linear function of T as is popularly assumed.
ROBERTJ. WIDLAR
National Semiconductor
Microcircuit Division
Santa Clara, Calif.

(n - 1)1 -k4T (

This expression gives VBEas a function of T for I , varying linearly with T.

To obtain VB,(T)for constant I,, a correction must be added.
The change inemitter-base voltage going fromone collector current to
a second is given by

REFEXENCES

Sah.

[I ] C. T.
EBea of surface recombinationand channel on P-N junction and transistor
characteristics, IRE Trans. on Elecrron Deoices, vol. ED-9, pp. 5&108, January 1%2.
[2 ] J. E.Iwersen, A. R.Bray, and J. J. Kleimack, Low-currentalpha in silicon transistors,
IRE Trans. on Electron Devices, vol. ED-9, pp. 474-478, November 1%2.
131 J. F. Gibbons and H. S. Horn, A circuit with logarithmic transfer response over 9
decades, IEEE Trans. on Circuit Theory, vol. (JT-1 I , pp. 37S384, September 1 9 6 4 .
[4] A. H. Hotfait and R. D. Thorton, Limitations of transistor DC amplifiers. Proc.
IEEE. vol. 52, pp. 17%184, February 1964.
[5] D. HiIbiber, A new semiconductor voltage standard, in 1% Inrernarl Solid Store
Circvirs Con/: Digcsr of Tech. Papers, vol. VII, p. 32.
[6] R. J.Widlar. Some circuit design techniques for linear integrated circuits, IEEE
Trans. on Circuit TFuory. vol. XII, pp. 586590, December 1%5.
171
A unique circuit design for a high performance operational amplifier especially
suited to monolithic construction, Proc. ofNEC, vol. XXI. pp. 8-9, October l%5.
[8] A. J. W. hi. Van Overbeek, Tunable resonant circuits suitable for integration, in
19x55 Internarl SolidStare Circuirs Con/: Digesr of Tech. Papers, vol. VIII, p. 92.
[9] R. J. Widlar, Core memory sense ampliljer designs using an integrated dual wrnparator, Fairchild Semiconductor APP-123, pp. 11-12.

-.